[Intel-gfx] [PATCH 15/19] drm: Update vblank->last in drm_update_vblank_count()

[Ville Syrjälä]

On Wed, Aug 06, 2014 at 03:08:25PM +0200, Daniel Vetter wrote:
> On Wed, Aug 06, 2014 at 02:49:58PM +0300, ville.syrjala at linux.intel.com wrote:
> > From: Ville Syrjälä <ville.syrjala at linux.intel.com>
> > 
> > We should update the last in drm_update_vblank_count() to avoid applying
> > the diff more than once. This could occur eg. if drm_vblank_off() gets
> > called multiple times for the crtc.
> > 
> > Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
> 
> Currently we update ->last when disabling the vblank and use it when
> re-enabling it. Those calls should be symmetric, except for driver bugs.
> Imo would be better to tighten up the checks for that.
> 
> Or do I completely misunderstand what's going on here?

The issue is that we want to call drm_vblank_off() in
intel_sanitize_crtc() for already disabled crtcs in order to
to bump the refcount and set inmodeset=1 so that
drm_vblank_get() won't try to enable vblank interrupts if someone calls
drm_vblank_get() on it. So during suspend+resume we can get two
drm_vblank_off() calls for the same crtc w/o a drm_vblank_on() in
between.

Although this patch doesn't actually fix the problem really since
vblank counter query during sanitize will just return 0 because the pipe
isn't active, so there's a later patch to just override .last=0 in
intel_sanitize_crtc(). Another approach might be to set .last=0 in
drm_vblank_off() itself (after vblank_disable_and_save()), mainly
just to hide the details from the caller.


> -Daniel
> 
> > ---
> >  drivers/gpu/drm/drm_irq.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
> > index 0523f5b..67507a4 100644
> > --- a/drivers/gpu/drm/drm_irq.c
> > +++ b/drivers/gpu/drm/drm_irq.c
> > @@ -109,6 +109,8 @@ static void drm_update_vblank_count(struct drm_device *dev, int crtc)
> >  	if (diff == 0)
> >  		return;
> >  
> > +	vblank->last = cur_vblank;
> > +
> >  	/* Reinitialize corresponding vblank timestamp if high-precision query
> >  	 * available. Skip this step if query unsupported or failed. Will
> >  	 * reinitialize delayed at next vblank interrupt in that case.
> > -- 
> > 1.8.5.5
> > 
> > _______________________________________________
> > Intel-gfx mailing list
> > Intel-gfx at lists.freedesktop.org
> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> 
> -- 
> Daniel Vetter
> Software Engineer, Intel Corporation
> +41 (0) 79 365 57 48 - http://blog.ffwll.ch

-- 
Ville Syrjälä
Intel OTC